When comparing the estimated values for the four CTMI
parameters with the values defined from all experimental data
available, Table 1 clearly shows that the Tmin, Topt, Tmax and mopt
derived from the four OED/PE experiments approximate these
values. Estimates obtained from the four equidistant experiments
significantly deviate and are mostly rather unrealistic.
For the equidistant experiments, the optimum growth temper-
ature is chosen between the third and the fourth temperature level
and the maximum growth temperature at the fourth temperature
level. This is due to the fact that not enough information is available
to obtain a realistic curvature of the CTMI model. Previous obser-
vations showed that temperature levels close to Tmax are required
for an accurate estimation (see Van Derlinden et al., 2008). This
implies that the a priori Tmax value must be rather accurate. The
unrealistic mopt estimates automatically follow from the inaccurate
Topt value.
In contrast, the third and fourth temperature level in the OED/PE
approach are both situated in the super-optimal temperature range.
The combination of these two levels with the second level (37 C),
which is a little below the optimum growth temperature yields
sufficient information for accurate CTMI identification.
Based on these two examples it can be stated that an OED/PE-
based design yields realistic CTMI parameter estimates, even
when only four experiments are performed. In addition, Bernaerts
et al. (2005) concluded that using an OED/PE design also leads to
much smaller confidence regions. These confidence regions can,
however, in the presented examples not be calculated as only four
experiments are performed to estimate four parameters.
When comparing the estimated values for the four CTMI
parameters with the values defined from all experimental data
available, Table 1 clearly shows that the Tmin, Topt, Tmax and mopt
derived from the four OED/PE experiments approximate these
values. Estimates obtained from the four equidistant experiments
significantly deviate and are mostly rather unrealistic.
For the equidistant experiments, the optimum growth temper-
ature is chosen between the third and the fourth temperature level
and the maximum growth temperature at the fourth temperature
level. This is due to the fact that not enough information is available
to obtain a realistic curvature of the CTMI model. Previous obser-
vations showed that temperature levels close to Tmax are required
for an accurate estimation (see Van Derlinden et al., 2008). This
implies that the a priori Tmax value must be rather accurate. The
unrealistic mopt estimates automatically follow from the inaccurate
Topt value.
In contrast, the third and fourth temperature level in the OED/PE
approach are both situated in the super-optimal temperature range.
The combination of these two levels with the second level (37 C),
which is a little below the optimum growth temperature yields
sufficient information for accurate CTMI identification.
Based on these two examples it can be stated that an OED/PE-
based design yields realistic CTMI parameter estimates, even
when only four experiments are performed. In addition, Bernaerts
et al. (2005) concluded that using an OED/PE design also leads to
much smaller confidence regions. These confidence regions can,
however, in the presented examples not be calculated as only four
experiments are performed to estimate four parameters.
การแปล กรุณารอสักครู่..
